DP/EP Support for gpt-oss with deepep-ht comm kernel on SM100 (#23608)
This commit is contained in:
@@ -897,6 +897,7 @@ class Fp8MoEMethod(FusedMoEMethodBase):
|
||||
self,
|
||||
prepare_finalize: FusedMoEPrepareAndFinalize,
|
||||
moe: FusedMoEConfig,
|
||||
layer: torch.nn.Module,
|
||||
) -> FusedMoEPermuteExpertsUnpermute:
|
||||
from vllm.model_executor.layers.fused_moe import (
|
||||
BatchedTritonOrDeepGemmExperts, TritonOrDeepGemmExperts)
|
||||
|
||||
Reference in New Issue
Block a user